Jan Hubicka [Mon, 16 Jan 2023 14:40:45 +0000 (15:40 +0100)]
Disable gather/scatter for zen4
this patch adds more tunes for zen4:
- new tunes for avx512 scater instructions.
In micro benchmarks these seems consistent loss compared to open-coded coe
- disable use of gather for zen4
While these are win for a micro benchmarks (based on TSVC), enabling gather
is a loss for parest. So for now it seems safe to keep it off.
- disable pass to avoid FMA chains for znver4 since fmadd was optimized and does not seem
to cause regressions.
* config/i386/i386.cc (ix86_vectorize_builtin_scatter): Guard scatter
by TARGET_USE_SCATTER.
* config/i386/i386.h (TARGET_USE_SCATTER_2PARTS,
TARGET_USE_SCATTER_4PARTS, TARGET_USE_SCATTER): New macros.
* config/i386/x86-tune.def (TARGET_USE_SCATTER_2PARTS,
TARGET_USE_SCATTER_4PARTS, TARGET_USE_SCATTER): New tunes.
(X86_TUNE_AVOID_256FMA_CHAINS, X86_TUNE_AVOID_512FMA_CHAINS): Disable
for znver4. (X86_TUNE_USE_GATHER): Disable for zen4.
arm: Split up MVE _Generic associations to prevent type clashes [PR107515]
With these previous patches:
https://gcc.gnu.org/pipermail/gcc-patches/2022-November/606586.html
https://gcc.gnu.org/pipermail/gcc-patches/2022-November/606587.html
we enabled the MVE overloaded _Generic associations to handle more
scalar types, however at PR 107515 we found a new regression that
wasn't detected in our testing:
With glibc's posix/types.h:
```
typedef signed int __int32_t;
...
typedef __int32_t int32_t;
```
We would get a `error: '_Generic' specifies two compatible types`
from `__ARM_mve_coerce3` because of `type: param`, when `type` is
`int` and `int32_t: param` both being the same under the hood.
The same did not happen with Newlib's header sys/_stdint.h:
```
typedef long int __int32_t;
...
typedef __int32_t int32_t ;
```
which worked fine, because it uses `long int`.
The same could feasibly happen in `__ARM_mve_coerce2` between
`__fp16` and `float16_t`.
The solution here is to break the _Generic down so that the similar
types don't appear at the same level, as is done in `__ARM_mve_typeid`
* g++.dg/tree-ssa/pr86544.C: Add .POPCOUNT to tree scan regex.
* gcc.dg/tree-ssa/popcount.c: Likewise.
* gcc.dg/tree-ssa/popcount2.c: Likewise.
* gcc.dg/tree-ssa/popcount3.c: Likewise.
* gcc.target/aarch64/popcount4.c: Likewise.
* gcc.target/i386/pr95771.c: Likewise, and...
* gcc.target/i386/pr95771-2.c: ...split int128 test from above,
since this would emit just a single IFN if a TI optab is added.
Andrew Carlotti [Thu, 22 Dec 2022 02:14:06 +0000 (02:14 +0000)]
Add c[lt]z idiom recognition
This recognises the patterns of the form:
while (n & 1) { n >>= 1 }
Unfortunately there are currently two issues relating to this patch.
Firstly, simplify_using_initial_conditions does not recognise that
(n != 0) and ((n & 1) == 0) implies that ((n >> 1) != 0).
This preconditions arise following the loop copy-header pass, and the
assumptions returned by number_of_iterations_exit_assumptions then
prevent final value replacement from using the niter result.
I'm not sure what is the best way to fix this - one approach could be to
modify simplify_using_initial_conditions to handle this sort of case,
but it seems that it basically wants the information that ranger could
give anway, so would something like that be a better option?
The second issue arises in the vectoriser, which is able to determine
that the niter->assumptions are always true.
When building with -march=armv8.4-a+sve -S -O3, we get this codegen:
foo (unsigned int b) {
int c = 0;
if (b == 0)
return PREC;
while (!(b & (1 << (PREC - 1)))) {
b <<= 1;
c++;
}
In essence, the vectoriser uses the niter information to determine
exactly how many iterations of the loop it needs to run. It then uses
SVE whilelo instructions to run this number of iterations. The original
loop counter is also vectorised, despite only being used in the final
iteration, and then the final value of this counter is used as the
return value (which is the same as the number of iterations it computed
in the first place).
This vectorisation is obviously bad, and I think it exposes a latent
bug in the vectoriser, rather than being an issue caused by this
specific patch.
gcc/ChangeLog:
* tree-ssa-loop-niter.cc (number_of_iterations_cltz): New.
(number_of_iterations_bitcount): Add call to the above.
(number_of_iterations_exit_assumptions): Add EQ_EXPR case for
c[lt]z idiom recognition.
gcc/testsuite/ChangeLog:
* gcc.dg/tree-ssa/cltz-max.c: New test.
* gcc.dg/tree-ssa/clz-char.c: New test.
* gcc.dg/tree-ssa/clz-int.c: New test.
* gcc.dg/tree-ssa/clz-long-long.c: New test.
* gcc.dg/tree-ssa/clz-long.c: New test.
* gcc.dg/tree-ssa/ctz-char.c: New test.
* gcc.dg/tree-ssa/ctz-int.c: New test.
* gcc.dg/tree-ssa/ctz-long-long.c: New test.
* gcc.dg/tree-ssa/ctz-long.c: New test.
Andrew Carlotti [Thu, 10 Nov 2022 15:56:51 +0000 (15:56 +0000)]
Add cltz_complement idiom recognition
This recognises patterns of the form:
while (n) { n >>= 1 }
This patch results in improved (but still suboptimal) codegen:
foo (unsigned int b) {
int c = 0;
while (b) {
b >>= 1;
c++;
}
return c;
}
foo:
.LFB11:
.cfi_startproc
cbz w0, .L3
clz w1, w0
tst x0, 1
mov w0, 32
sub w0, w0, w1
csel w0, w0, wzr, ne
ret
The conditional is unnecessary. phiopt could recognise a redundant csel
(using cond_removal_in_builtin_zero_pattern) when one of the inputs is a
clz call, but it cannot recognise the redunancy when the input is (e.g.)
(32 - clz).
I could perhaps extend this function to recognise this pattern in a later
patch, if this is a good place to recognise more patterns.
gcc/ChangeLog:
PR tree-optimization/94793
* tree-scalar-evolution.cc (expression_expensive_p): Add checks
for c[lt]z optabs.
* tree-ssa-loop-niter.cc (build_cltz_expr): New.
(number_of_iterations_cltz_complement): New.
(number_of_iterations_bitcount): Add call to the above.
gcc/testsuite/ChangeLog:
* lib/target-supports.exp (check_effective_target_clz)
(check_effective_target_clzl, check_effective_target_clzll)
(check_effective_target_ctz, check_effective_target_clzl)
(check_effective_target_ctzll): New.
* gcc.dg/tree-ssa/cltz-complement-max.c: New test.
* gcc.dg/tree-ssa/clz-complement-char.c: New test.
* gcc.dg/tree-ssa/clz-complement-int.c: New test.
* gcc.dg/tree-ssa/clz-complement-long-long.c: New test.
* gcc.dg/tree-ssa/clz-complement-long.c: New test.
* gcc.dg/tree-ssa/ctz-complement-char.c: New test.
* gcc.dg/tree-ssa/ctz-complement-int.c: New test.
* gcc.dg/tree-ssa/ctz-complement-long-long.c: New test.
* gcc.dg/tree-ssa/ctz-complement-long.c: New test.
Jonathan Wakely [Mon, 16 Jan 2023 10:15:41 +0000 (10:15 +0000)]
libstdc++: Fix copyright notice to use usual form [PR108413]
libstdc++-v3/ChangeLog:
PR libstdc++/108413
* include/c_compatibility/stdatomic.h: Change copyright line to
be consistent with other headers contributed under DCO terms.
* include/std/expected: Add full stop to copyright line.
* src/c++20/tzdb.cc: Likewise.
Jonathan Wakely [Fri, 13 Jan 2023 13:13:55 +0000 (13:13 +0000)]
libatomic: Use config/mingw/lock.c for --enable-threads=single
Without this change bootstrap fails for x86_64-w64-mingw32 with
--disable-threads=single because there is no lock.c file chosen by
libatomic's configure.
libatomic/ChangeLog:
* configure.tgt (config_path) [target_thread_file=single]:
Use 'mingw' config.
Jakub Jelinek [Mon, 16 Jan 2023 10:03:30 +0000 (11:03 +0100)]
contrib: Partial fix for failed update-copyright --this year [PR108413]
As mentioned on IRC or in PR108413, the last update-copyright.py --this year
failed and that is why we are in a strange state where some copyrights have
been updated and others have not.
The full list of errors I got was I think:
gcc/m2/mc-boot/GmcOptions.c: unrecognised copyright: comment (f, (const char *) "Copyright (C) ''2021'' Free Software Foundation, Inc.", 53);
gcc/m2/mc-boot/GmcOptions.c: unrecognised copyright: comment (f, (const char *) "Copyright (C) ''2021'' Free Software Foundation, Inc.", 53);
gcc/testsuite/gm2/switches/pedantic-params/pass/Strings.mod: unrecognised copyright holder: Faculty of Information Technology,
gcc/testsuite/gm2/switches/pedantic-params/pass/Strings2.mod: unrecognised copyright holder: Faculty of Information Technology,
libphobos/libdruntime/__builtins.di: unrecognised copyright: * Copyright: Copyright Digital Mars 2022
libstdc++-v3/src/c++17/fast_float/fast_float.h: unrecognised copyright holder: The fast_float authors
libstdc++-v3/include/c_compatibility/stdatomic.h: unrecognised copyright holder: The GCC developers
The following patch deals with the gcc/testsuite/gm2 ones and
with the fast_float.h one, ok for trunk?
Not really sure what we should do in the GmcOptions.c case
(perhaps obfuscate it in the source somehow by splitting
the string literals into different substrings
Perhaps "Copy" "right (" "C) ''..." would do it? Or do we want
to bump there each year (manually or by the script)?
E.g. in gcc.cc we have
printf ("Copyright %s 2023 Free Software Foundation, Inc.\n",
_("(C)"));
which also prints (C) nicer in Unicode if possible and is updated
by hand each year.
I have no idea about the libphobos case, we have tons of
libphobos/src/std/format/spec.d:Copyright: Copyright The D Language Foundation 2000-2013.
libphobos/src/std/random.d:Copyright: Copyright Andrei Alexandrescu 2008 - 2009, Joseph Rushton Wakeling 2012.
etc. lines and those aren't reported as errors.
And the last one is that I think for The GCC developers we should treat it
similarly like FSF and bump copyright on it.
Would
canon_gcc = 'The GCC developers'
self.add_package_author ('The GCC developers', canon_gcc)
self.add_package_author ('The GCC Developers', canon_gcc)
or something similar do the trick?
2023-01-16 Jakub Jelinek <jakub@redhat.com>
PR other/108413
* update-copyright.py (TestsuiteFilter): Add .mod and .rs extensions.
(GCCCopyright): Add 'The fast_float authors' as external author.
Jakub Jelinek [Mon, 16 Jan 2023 08:41:38 +0000 (09:41 +0100)]
x86: Avoid -Wuninitialized warnings on _mm*_undefined_* in C++ [PR105593]
In https://gcc.gnu.org/pipermail/gcc-patches/2023-January/609844.html
I've posted a patch to allow ignoring -Winit-self using GCC diagnostic
pragmas, such that one can mark self-initialization as intentional
disabling of -Wuninitialized warnings.
The following incremental patch uses that in the x86 intrinsic
headers.
Jakub Jelinek [Mon, 16 Jan 2023 08:40:14 +0000 (09:40 +0100)]
c, c++: Allow ignoring -Winit-self through pragmas [PR105593]
As mentioned in the PR, various x86 intrinsics need to return
an uninitialized vector. Currently they use self initialization
to avoid -Wuninitialized warnings, which works fine in C, but
doesn't work in C++ where -Winit-self is enabled in -Wall.
We don't have an attribute to mark a variable as knowingly
uninitialized (the uninitialized attribute exists but means
something else, only in the -ftrivial-auto-var-init context),
and trying to suppress either -Wuninitialized or -Winit-self
inside of the _mm_undefined_ps etc. intrinsic definitions
doesn't work, one needs to currently disable through pragmas
-Wuninitialized warning at the point where _mm_undefined_ps etc.
result is actually used, but that goes against the intent of
those intrinsics.
The -Winit-self warning option actually doesn't do any warning,
all we do is record a suppression for -Winit-self if !warn_init_self
on the decl definition and later look that up in uninit pass.
The following patch changes those !warn_init_self tests which
are true only based on the command line option setting, not based
on GCC diagnostic pragma overrides to
!warning_enabled_at (DECL_SOURCE_LOCATION (decl), OPT_Winit_self)
such that it takes them into account.
2023-01-16 Jakub Jelinek <jakub@redhat.com>
PR c++/105593
gcc/c/
* c-parser.cc (c_parser_initializer): Check warning_enabled_at
at the DECL_SOURCE_LOCATION (decl) for OPT_Winit_self instead
of warn_init_self.
gcc/cp/
* decl.cc (cp_finish_decl): Check warning_enabled_at
at the DECL_SOURCE_LOCATION (decl) for OPT_Winit_self instead
of warn_init_self.
gcc/testsuite/
* c-c++-common/Winit-self3.c: New test.
* c-c++-common/Winit-self4.c: New test.
* c-c++-common/Winit-self5.c: New test.
Kewen Lin [Mon, 16 Jan 2023 08:15:39 +0000 (02:15 -0600)]
rs6000: Teach rs6000_opaque_type_invalid_use_p about inline asm [PR108272]
As PR108272 shows, there are some invalid uses of MMA opaque
types in inline asm statements. This patch is to teach the
function rs6000_opaque_type_invalid_use_p for inline asm,
check and error any invalid use of MMA opaque types in input
and output operands.
PR target/108272
gcc/ChangeLog:
* config/rs6000/rs6000.cc (rs6000_opaque_type_invalid_use_p): Add the
support for invalid uses in inline asm, factor out the checking and
erroring to lambda function check_and_error_invalid_use.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/pr108272-1.c: New test.
* gcc.target/powerpc/pr108272-2.c: New test.
* gcc.target/powerpc/pr108272-3.c: New test.
* gcc.target/powerpc/pr108272-4.c: New test.
Aldy Hernandez [Tue, 10 Jan 2023 09:57:16 +0000 (10:57 +0100)]
[PR107608] [range-ops] Avoid folding into INF when flag_trapping_math.
As discussed in the PR, for trapping math, do not fold overflowing
operations into +-INF as doing so could elide a trap.
There is a minor adjustment to known_isinf() where it was mistakenly
returning true for an [infinity U NAN], whereas it should only return
true when the range is exclusively +INF or -INF. This is benign, as
there were no users of known_isinf up to now.
Tested on x86-64 Linux.
I also ran the glibc testsuite (git sources) on x86-64 and this patch
fixes:
* range-op-float.cc (range_operator_float::fold_range): Avoid
folding into INF when flag_trapping_math.
* value-range.h (frange::known_isinf): Return false for possible NANs.
Gaius Mulley [Sun, 15 Jan 2023 13:13:40 +0000 (13:13 +0000)]
Bugfix to allow testsuite/gm2/pim/pass/arraybool.mod to compile on ppc64le
This bug is exposed on the ppc64le platform. The expression
parser P3Build.bnf (and PHBuild.bnf) BuiltNot omitted to record
the current token position on the quad stack. The patch changes
all occurances of NEW to newBoolFrame to ensure that the tokenno
recorded in the bool frame is set to a sensible value.
BuildNot is fixed and improved to generate a virtual token
recording the position of the subexpression.
gcc/m2/ChangeLog:
* gm2-compiler/M2LexBuf.mod (isSrcToken): Add block comment.
Remove dead code.
* gm2-compiler/M2Quads.def (BuildNot): Add notTokPos parameter.
* gm2-compiler/M2Quads.mod (BuildNot): Add notTokPos parameter.
Create and push virtual token.
(PopBooltok): New procedure.
(PushBooltok): New procedure.
(PushBool): Re-implement using PushBooltok.
(PopBool): Re-implement using PopBooltok.
* gm2-compiler/P3Build.bnf (ConstFactor): Record token
position of NOT.
(Factor): Record token position of NOT.
* gm2-compiler/PHBuild.bnf (ConstFactor): Record token
position of NOT.
(Relation): Push token position.
(UnaryOrConstTerm): Push token position.
(AddOperator): Push token position.
(MulOperator): Push token position.
Jonathan Wakely [Sun, 15 Jan 2023 12:38:31 +0000 (12:38 +0000)]
libstdc++: Fix narrowing conversion in std/time/clock/utc/io.cc
For a port with signed char and unsigned wchar_t initializing a wchar_t
array with a char is a narrowing conversion. The code is wrong for
assuming that (int)'a' == (int)L'a' anyway, so fix it properly by using
ctype<wchar_t>::widen(char).
libstdc++-v3/ChangeLog:
* testsuite/std/time/clock/utc/io.cc: Use ctype to widen char.
Björn Schäpers [Tue, 13 Dec 2022 21:02:47 +0000 (22:02 +0100)]
libstdc++: Deliver names of C functions in <stacktrace>
__cxa_demangle is only to demangle C++ names, for all C functions,
extern "C" functions, and including main it returns -2, in that case
just adapt the given name. Otherwise it's kept empty, which doesn't look
nice in the stacktrace.
libstdc++-v3/ChangeLog:
* include/std/stacktrace (stacktrace_entry::_S_demangle): Use
raw __name if __cxa_demangle could not demangle it.
Jonathan Wakely [Sat, 14 Jan 2023 13:39:48 +0000 (13:39 +0000)]
libstdc++: Disable unwanted code for --with-libstdcxx-zoneinfo=no
This allows most of the tzdb functionality to be disabled by
configuring with --with-libstdcxx-zoneinfo=no. This might be desirable
for small targets that don't need the time zone support.
libstdc++-v3/ChangeLog:
* src/c++20/tzdb.cc (TZDB_DISABLED): Disable all code for
loading tzdb.
* testsuite/std/time/tzdb/leap_seconds.cc: Require tzdb
effective target.
* testsuite/std/time/tzdb_list/1.cc: Likewise.
Jonathan Wakely [Sat, 14 Jan 2023 13:33:58 +0000 (13:33 +0000)]
libstdc++: Embed a static copy of tzdata.zi
This adds a copy of the tzdata.zi file to the library, and allows
configuring to use it instead of a copy read from disk at runtime.
The content of the file is in the public domain, but will need to be
updated to the latest upstream file before making GCC releases.
libstdc++-v3/ChangeLog:
* acinclude.m4 (GLIBCXX_ZONEINFO_DIR): Replace the
--with-libstdcxx-zoneinfo-dir configure option with
--with-libstdcxx-zoneinfo with yes/no/static choices as well as
a directory.
* config.h.in: Regenerate.
* configure: Regenerate.
* doc/xml/manual/configure.xml: Document configure option.
* doc/html/manual/configure.html: Regenerate.
* src/c++20/Makefile.am: Generate tzdata.zi.h header.
* src/c++20/Makefile.in: Regenerate.
* src/c++20/tzdb.cc (__gnu_cxx::zoneinfo_dir_override): Return a
null pointer if no directory is configured.
(zoneinfo_dir): Replace with ...
(zoneinfo_file): New function.
(tzdata_stream): New istream class.
(remote_version, reload_tzdb): Use tzdata_stream.
* testsuite/lib/libstdc++.exp (check_effective_target_tzdb):
Check new _GLIBCXX_STATIC_TZDATA macro and ignore presence of
tzdata.zi file in default location.
* src/c++20/tzdata.zi: New file.
In the before-IRA era, ORDER_REGS_FOR_LOCAL_ALLOC was called for each
function in Xtensa, and there was register allocation table reordering
for leaf functions to compensate for the poor performance of local-alloc.
Today the adjustment hook is still called via its alternative
ADJUST_REG_ALLOC_ORDER, but it is only called once at the start of the IRA,
and leaf_function_p() erroneously returns true and also gives no argument
count.
That straightforwardly misleads register allocation that all functions are
always leaves with no arguments, which leads to inefficiencies in allocation
results.
Fortunately, IRA is smart enough than local-alloc to not need such assistance.
This patch does away with the antiquated by removing the wreckage that no
longer works.
gcc/ChangeLog:
* config/xtensa/xtensa-protos.h (order_regs_for_local_alloc):
Rename to xtensa_adjust_reg_alloc_order.
* config/xtensa/xtensa.cc (xtensa_adjust_reg_alloc_order):
Ditto. And also remove code to reorder register numbers for
leaf functions, rename the tables, and adjust the allocation
order for the call0 ABI to use register A0 more.
(xtensa_leaf_regs): Remove.
* config/xtensa/xtensa.h (REG_ALLOC_ORDER): Cosmetics.
(order_regs_for_local_alloc): Rename as the above.
(LEAF_REGISTERS, LEAF_REG_REMAP, leaf_function): Remove.
Jonathan Wakely [Sat, 14 Jan 2023 15:41:58 +0000 (15:41 +0000)]
libstdc++: Fix ostream insertion operators for calendar types
libstdc++-v3/ChangeLog:
* include/bits/chrono_io.h (operator<<): Fix syntax errors.
* testsuite/std/time/month_day/io.cc: New test.
* testsuite/std/time/month_day_last/io.cc: New test.
* testsuite/std/time/month_weekday/io.cc: New test.
* testsuite/std/time/month_weekday_last/io.cc: New test.
* testsuite/std/time/weekday_indexed/io.cc: New test.
* testsuite/std/time/weekday_last/io.cc: New test.
* testsuite/std/time/year_month/io.cc: New test.
* testsuite/std/time/year_month_day_last/io.cc: New test.
* testsuite/std/time/year_month_weekday/io.cc: New test.
* testsuite/std/time/year_month_weekday_last/io.cc: New test.
Jakub Jelinek [Sat, 14 Jan 2023 09:17:14 +0000 (10:17 +0100)]
c++: Avoid incorrect shortening of divisions [PR108365]
The following testcase is miscompiled, because we shorten the division
in a case where it should not be shortened.
Divisions (and modulos) can be shortened if it is unsigned division/modulo,
or if it is signed division/modulo where we can prove the dividend will
not be the minimum signed value or divisor will not be -1, because e.g.
on sizeof(long long)==sizeof(int)*2 && __INT_MAX__ == 0x7fffffff targets
(-2147483647 - 1) / -1 is UB
but
(int) (-2147483648LL / -1LL) is not, it is -2147483648.
The primary aim of both the C and C++ FE division/modulo shortening I assume
was for the implicit integral promotions of {,signed,unsigned} {char,short}
and because at this point we have no VRP information etc., the shortening
is done if the integral promotion is from unsigned type for the divisor
or if the dividend is an integer constant other than -1.
This works fine for char/short -> int promotions when char/short have
smaller precision than int - unsigned char -> int or unsigned short -> int
will always be a positive int, so never the most negative.
Now, the C FE checks whether orig_op0 is TYPE_UNSIGNED where op0 is either
the same as orig_op0 or that promoted to int, I think that works fine,
if it isn't promoted, either the division/modulo common type will have the
same precision as op0 but then the division/modulo is unsigned and so
without UB, or it will be done in wider precision (e.g. because op1 has
wider precision), but then op0 can't be minimum signed value. Or it has
been promoted to int, but in that case it was again from narrower type and
so never minimum signed int.
But the C++ FE was checking if op0 is a NOP_EXPR from TYPE_UNSIGNED.
First of all, not sure if the operand of NOP_EXPR couldn't be non-integral
type where TYPE_UNSIGNED wouldn't be meaningful, but more importantly,
even if it is a cast from unsigned integral type, we only know it can't be
minimum signed value if it is a widening cast, if it is same precision or
narrowing cast, we know nothing.
So, the following patch for the NOP_EXPR cases checks just in case that
it is from integral type and more importantly checks it is a widening
conversion, and then next to it also allows op0 to be just unsigned,
promoted or not, as that is what the C FE will do for those cases too
and I believe it must work - either the division/modulo common type
will be that unsigned type, then we can shorten and don't need to worry
about UB, or it will be some wider signed type but then it can't be most
negative value of the wider type.
And changes both the C and C++ FEs to do the same thing, using a helper
function in c-family.
2023-01-14 Jakub Jelinek <jakub@redhat.com>
PR c++/108365
* c-common.h (may_shorten_divmod): New static inline function.
* c-typeck.cc (build_binary_op): Use may_shorten_divmod for integral
division or modulo.
* typeck.cc (cp_build_binary_op): Use may_shorten_divmod for integral
division or modulo.
* c-c++-common/pr108365.c: New test.
* g++.dg/opt/pr108365.C: New test.
* g++.dg/warn/pr108365.C: New test.
Alexandre Oliva [Sat, 14 Jan 2023 00:15:44 +0000 (21:15 -0300)]
hash table: enforce testing is_empty before is_deleted
Existing hash_table traits that use the same representation for empty
and deleted slots reject marking slots as deleted, and to not pass
is_deleted for slots that pass is_empty.
Nevertheless, nearly everywhere, we only test for is_deleted after
checking that !is_empty first. The one exception was the copy
constructor, that would fail if traits recognized is_empty slots as
is_deleted, but then refused to mark_deleted.
This asymmetry is neither necessary nor desirable, and there is a
theoretical risk that traits might not only fail to refuse to
mark_deleted, but also return is_deleted for is_empty slots.
This patch introduces checks that detect these potentially problematic
situations, and reorders the tests in the copy constructor so as to
use the conventional testing order and thus avoid them.
for gcc/ChangeLog
* hash-table.h (is_deleted): Precheck !is_empty.
(mark_deleted): Postcheck !is_empty.
(copy constructor): Test is_empty before is_deleted.
Alexandre Oliva [Sat, 14 Jan 2023 00:15:42 +0000 (21:15 -0300)]
[PR42093] [arm] [thumb2] disable tree-dce for test
CD-DCE introduces blocks to share common PHI nodes, which replaces a
backwards branch that used to prevent the thumb2 jump table shortening
that PR42093 tested for. In order to keep on testing that the
backward branch prevents the jumptable shortening, disable tree-dce.
Alexandre Oliva [Sat, 14 Jan 2023 00:15:41 +0000 (21:15 -0300)]
[PR40457] [arm] expand SI-aligned movdi into pair of movsi
When expanding a misaligned DImode move, emit aligned SImode moves if
the parts are sufficiently aligned. This enables neighboring stores
to be peephole-combined into stm, as expected by the PR40457 testcase,
even after SLP vectorizes the originally aligned SImode stores into a
misaligned DImode store.
David Malcolm [Fri, 13 Jan 2023 22:51:26 +0000 (17:51 -0500)]
analyzer: add heuristics for switch on enum type [PR105273]
Assume that switch on an enum doesn't follow an implicit default
skipping all cases when all enum values are covered by cases.
Fixes various false positives from -Wanalyzer-use-of-uninitialized-value
such as this one seen in Doom:
p_maputl.c: In function 'P_BoxOnLineSide':
p_maputl.c:151:8: warning: use of uninitialized value 'p1' [CWE-457] [-Wanalyzer-use-of-uninitialized-value]
151 | if (p1 == p2)
| ^
'P_BoxOnLineSide': events 1-5
|
| 115 | int p1;
| | ^~
| | |
| | (1) region created on stack here
| | (2) capacity: 4 bytes
|......
| 118 | switch (ld->slopetype)
| | ~~~~~~
| | |
| | (3) following 'default:' branch...
|......
| 151 | if (p1 == p2)
| | ~
| | |
| | (4) ...to here
| | (5) use of uninitialized value 'p1' here
|
where "ld->slopetype" is a "slopetype_t" enum, and for every value of
that enum the switch has a case that initializes "p1".
gcc/analyzer/ChangeLog:
PR analyzer/105273
* region-model.cc (has_nondefault_case_for_value_p): New.
(has_nondefault_cases_for_all_enum_values_p): New.
(region_model::apply_constraints_for_gswitch): Skip
implicitly-created "default" when switching on an enum
and all enum values have non-default cases.
(rejected_default_case::dump_to_pp): New.
* region-model.h (region_model_context::possibly_tainted_p): New
decl.
(class rejected_default_case): New.
* sm-taint.cc (region_model_context::possibly_tainted_p): New.
* supergraph.cc (switch_cfg_superedge::dump_label_to_pp): Dump
when implicitly_created_default_p.
(switch_cfg_superedge::implicitly_created_default_p): New.
* supergraph.h
(switch_cfg_superedge::implicitly_created_default_p): New decl.
gcc/testsuite/ChangeLog:
PR analyzer/105273
* gcc.dg/analyzer/switch-enum-1.c: New test.
* gcc.dg/analyzer/switch-enum-2.c: New test.
* gcc.dg/analyzer/switch-enum-pr105273-git-vreportf-2.c: New test.
* gcc.dg/analyzer/switch-enum-taint-1.c: New test.
* gcc.dg/analyzer/switch-wrong-enum.c: New test.
* gcc.dg/analyzer/torture/switch-enum-pr105273-doom-p_floor.c: New
test.
* gcc.dg/analyzer/torture/switch-enum-pr105273-doom-p_maputl.c:
New test.
* gcc.dg/analyzer/torture/switch-enum-pr105273-git-vreportf-1.c:
New test.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Eric Botcazou [Fri, 13 Jan 2023 21:11:32 +0000 (22:11 +0100)]
Small fix for -fdump-ada-spec
This is needed to support the _Float32 and _Float64 types.
gcc/c-family/
* c-ada-spec.cc (is_float32): New function.
(is_float64): Likewise.
(is_float128): Tweak.
(dump_ada_node) <REAL_TYPE>: Call them to recognize more types.
Eric Botcazou [Fri, 13 Jan 2023 21:05:28 +0000 (22:05 +0100)]
Fix PR rtl-optimization/108274
Unlike other IPA passes, the ICF pass can be run at -O0 and some testcases
rely on this in the testsuite. Now it effectively creates a tail call so
the DF information needs be updated in this case after epilogue creation.
gcc/
PR rtl-optimization/108274
* function.cc (thread_prologue_and_epilogue_insns): Also update the
DF information for calls in a few more cases.
* gm2-gcc/m2options.h (M2Options_SetVerbose): Export the
function.
* gm2-lang.cc: Handle OPT_v, passing it to the compiler.
* lang-specs.h: Pass -v to cc1gm2.
This change updates the atomic libcall support to fix the following
issues:
1) A internal compiler error with -fno-sync-libcalls.
2) When sync libcalls are disabled, we don't generate libcalls for
libatomic.
3) There is no sync libcall support for targets other than linux.
As a result, non-atomic stores are silently emitted for types
smaller or equal to the word size. There are now a few atomic
libcalls in the libgcc code, so we need sync support on all
targets.
2023-01-13 John David Anglin <danglin@gcc.gnu.org>
gcc/ChangeLog:
* config/pa/pa-linux.h (TARGET_SYNC_LIBCALL): Delete define.
* config/pa/pa.cc (pa_init_libfuncs): Use MAX_SYNC_LIBFUNC_SIZE
define.
* config/pa/pa.h (TARGET_SYNC_LIBCALLS): Use flag_sync_libcalls.
(MAX_SYNC_LIBFUNC_SIZE): Define.
(TARGET_CPU_CPP_BUILTINS): Define __SOFTFP__ when soft float is
enabled.
* config/pa/pa.md (atomic_storeqi): Emit __atomic_exchange_1
libcall when sync libcalls are disabled.
(atomic_storehi, atomic_storesi, atomic_storedi): Likewise.
(atomic_loaddi): Emit __atomic_load_8 libcall when sync libcalls
are disabled on 32-bit target.
* config/pa/pa.opt (matomic-libcalls): New option.
* doc/invoke.texi (HPPA Options): Update.
libgcc/ChangeLog:
* config.host (hppa*64*-*-linux*): Adjust tmake_file to use
pa/t-pa64-linux.
(hppa*64*-*-hpux11*): Adjust tmake_file to use pa/t-pa64-hpux
instead of pa/t-hpux and pa/t-pa64.
* config/pa/linux-atomic.c: Define u32 type.
(ATOMIC_LOAD): Define new macro to implement atomic_load_1,
atomic_load_2, atomic_load_4 and atomic_load_8. Update sync
defines to use atomic_load calls for type.
(SYNC_LOCK_LOAD_2): New macro to implement __sync_lock_load_8.
* config/pa/sync-libfuncs.c: New file.
* config/pa/t-netbsd (LIB2ADD_ST): Define.
* config/pa/t-openbsd (LIB2ADD_ST): Define.
* config/pa/t-pa64-hpux: New file.
* config/pa/t-pa64-linux: New file.
sched-deps: do not schedule pseudos across calls [PR108117]
Scheduling across calls in the pre-RA scheduler is problematic: we do
not take liveness info into account, and are thus prone to extending
lifetime of a pseudo over the loop, requiring a callee-saved hardreg
or causing a spill.
If current function called a setjmp, lifting an assignment over a call
may be incorrect if a longjmp would happen before the assignment.
Thanks to Jose Marchesi for testing on AArch64.
gcc/ChangeLog:
PR rtl-optimization/108117
PR rtl-optimization/108132
* sched-deps.cc (deps_analyze_insn): Do not schedule across
calls before reload.
gcc/testsuite/ChangeLog:
PR rtl-optimization/108117
PR rtl-optimization/108132
* gcc.dg/pr108117.c: New test.
Jakub Jelinek [Fri, 13 Jan 2023 17:23:57 +0000 (18:23 +0100)]
c++: Avoid some false positive -Wfloat-conversion warnings with extended precision [PR108285]
On the following testcase trunk emits a false positive warning on ia32.
convert_like_internal is there called with type of double and
expr EXCESS_PRECISION_EXPR with float type with long double operand
2.L * (long double) x.
Now, for the code generation we do the right thing, cp_convert
to double from that 2.L * (long double) x, but we call even
cp_convert_and_check with that and that emits the -Wfloat-conversion
warning. Looking at what the C FE does in this case, it calls
convert_and_check with the EXCESS_PRECISION_EXPR expression rather
than its operand, and essentially uses the operand for code generation
and EXCESS_PRECISION_EXPR itself for warnings.
The following patch does that too for the C++ FE.
2023-01-13 Jakub Jelinek <jakub@redhat.com>
PR c++/108285
* cvt.cc (cp_convert_and_check): For EXCESS_PRECISION_EXPR
use its operand except that for warning purposes use the original
EXCESS_PRECISION_EXPR.
* call.cc (convert_like_internal): Only look through
EXCESS_PRECISION_EXPR when calling cp_convert, not when calling
cp_convert_and_check.
Gaius Mulley [Fri, 13 Jan 2023 17:17:36 +0000 (17:17 +0000)]
Recalibrate the timeouts for the larger code tests
Some of the larger code tests timeout when -O3 is given.
This patch increase the map and pimlib-base-run-pass tests.
gcc/testsuite/ChangeLog:
* gm2/examples/map/pass/examples-map-pass.exp:
Call gm2_push_timeout 30 before foreach testcase.
Call gm2_pop_timeout after the foreach statement.
* gm2/pimlib/base/run/pass/pimlib-base-run-pass.exp:
Call gm2_push_timeout 20 before foreach testcase.
Call gm2_pop_timeout after the foreach statement.
Jakub Jelinek [Fri, 13 Jan 2023 16:37:52 +0000 (17:37 +0100)]
testsuite: Add another testcase from PR107131
This one is hand reduced to problematic code from optimized dump
that used to be miscompiled during combine starting with
r12-303 and fixed with r13-3530 aka PR107172 fix.
2023-01-13 Jakub Jelinek <jakub@redhat.com>
PR target/107131
* gcc.c-torture/execute/pr107131.c: New test.
This patch adds cde feature (optional) support for Cortex-M55 CPU, please refer
[1] for more details. To use this feature we need to specify +cdecpN
(e.g. -mcpu=cortex-m55+cdecp<N>), where N is the coprocessor number 0 to 7.
Qing Zhao [Fri, 13 Jan 2023 15:08:00 +0000 (15:08 +0000)]
Replace flag_strict_flex_arrays with DECL_NOT_FLEXARRAY in middle-end.
We should not directly check flag_strict_flex_arrays in the
middle end. Instead, check DECL_NOT_FLEXARRAY(array_field_decl) which is set
by C/C++ FEs according to -fstrict-flex-arrays and the corresponding
attribute attached to the array_field.
As a result, We will lose the LEVEL information of -fstrict-flex-arrays in
the middle end. -Wstrict-flex-arrays will not be able to issue such
information. update the testing cases accordingly.
gcc/ChangeLog:
* attribs.cc (strict_flex_array_level_of): Move this function to ...
* attribs.h (strict_flex_array_level_of): Remove the declaration.
* gimple-array-bounds.cc (array_bounds_checker::check_array_ref):
replace the referece to strict_flex_array_level_of with
DECL_NOT_FLEXARRAY.
* tree.cc (component_ref_size): Likewise.
Arsen Arsenović [Tue, 10 Jan 2023 10:03:05 +0000 (11:03 +0100)]
libstdc++: Fix a few !HOSTED test regressions
libstdc++-v3/ChangeLog:
* testsuite/20_util/to_chars/version.cc: Mark hosted-only.
* testsuite/20_util/uses_allocator/lwg3677.cc: Ditto.
* testsuite/20_util/weak_ptr/cons/self_move.cc: Ditto.
* testsuite/std/ranges/adaptors/as_rvalue/1.cc: Replace usage of
std::make_unique with a freestanding-compatible wrapper around
unique_ptr.
* testsuite/21_strings/basic_string_view/operations/contains/char.cc:
Don't test for presence of __cpp_lib_string_contains on !HOSTED.
* testsuite/21_strings/basic_string_view/operations/contains/char/2.cc:
Ditto.
* testsuite/std/ranges/version_c++23.cc: Don't test for presence
of __cpp_lib_ranges in !HOSTED.
Dimitrij Mijoski [Tue, 10 Jan 2023 12:58:59 +0000 (13:58 +0100)]
libstdc++: Fix Unicode codecvt and add tests [PR86419]
Fixes the conversion from UTF-8 to UTF-16 to properly return partial
instead ok.
Fixes the conversion from UTF-16 to UTF-8 to properly return partial
instead ok.
Fixes the conversion from UTF-8 to UCS-2 to properly return partial
instead error.
Fixes the conversion from UTF-8 to UCS-2 to treat 4-byte UTF-8 sequences
as error just by seeing the leading byte.
Fixes UTF-8 decoding for all codecvts so they detect error at the end of
the input range when the last code point is also incomplete.
libstdc++-v3/ChangeLog:
PR libstdc++/86419
* src/c++11/codecvt.cc (read_utf8_code_point): Correctly detect
errors in incomplete multibyte sequences.
(utf16_in): Remove surrogates parameter. Fix conditions for
returning partial.
(utf16_out): Fix condition for returning partial.
(ucs2_in): Do not pass surrogates argument to utf16_in.
* testsuite/22_locale/codecvt/codecvt_unicode.cc: New test.
* testsuite/22_locale/codecvt/codecvt_unicode.h: New header for
tests.
* testsuite/22_locale/codecvt/codecvt_unicode_wchar_t.cc: New
test.
Jonathan Wakely [Thu, 12 Jan 2023 10:40:49 +0000 (10:40 +0000)]
libstdc++: Fix unintended layout change to std::basic_filebuf [PR108331]
GCC 13 has a new implementation of gthr-win32.h which supports C++11
mutexes, threads etc. but this causes an unintended ABI break. The
__gthread_mutex_t type is always used in std::basic_filebuf even in
C++98, so independent of whether C++11 sync primitives work or not.
Because that type changed for the win32 thread model, we have a layout
change in std::basic_filebuf. The member is completely unused, it just
gets passed to the std::__basic_file constructor and ignored. So we
don't need that mutex to actually work, we just need its layout to not
change.
Introduce a new __gthr_win32_legacy_mutex_t struct in gthr-win32.h with
the old layout, and conditionally use that in std::basic_filebuf.
PR libstdc++/108331
libgcc/ChangeLog:
* config/i386/gthr-win32.h (__gthr_win32_legacy_mutex_t): New
struct matching the previous __gthread_mutex_t struct.
(__GTHREAD_LEGACY_MUTEX_T): Define.
libstdc++-v3/ChangeLog:
* config/io/c_io_stdio.h (__c_lock): Define as a typedef for
__GTHREAD_LEGACY_MUTEX_T if defined.
aarch64: Fix DWARF frame register sizes for predicates
Jakub pointed out that __builtin_init_dwarf_reg_size_table
set the size of predicate registers to their current runtime
size when compiled with +sve, but to 8 bytes otherwise.
As explained in the comment, both behaviours are wrong.
Predicates change size with VL and should never need to be
restored during unwinding.
In contrast, the call-saved FP&SIMD frame registers are 8 bytes
(even though the hardware registers are at least 16 bytes) and
the call-clobbered registers have zero size. A zero size seems
correct for predicates too.
gcc/
* config/aarch64/aarch64.cc (aarch64_dwarf_frame_reg_mode): New
function.
(TARGET_DWARF_FRAME_REG_MODE): Define.
gcc/testsuite/
* gcc.target/aarch64/dwarf_reg_size_1.c: New test.
* gcc.target/aarch64/dwarf_reg_size_2.c: Likewise.
Richard Biener [Fri, 13 Jan 2023 10:01:32 +0000 (10:01 +0000)]
aarch64: Don't update EH info when folding [PR107209]
The AArch64 folders tried to update EH info on the fly, bypassing
the folder's attempts to remove dead EH edges later. This triggered
an ICE when folding a potentially-trapping call to a constant.
gcc/
PR target/107209
* config/aarch64/aarch64.cc (aarch64_gimple_fold_builtin): Don't
update EH info on the fly.
gcc/testsuite/
* gcc.target/aarch64/pr107209.c: New test.
Co-Authored-By: Richard Biener <rguenther@suse.de>
Gaius Mulley [Fri, 13 Jan 2023 01:08:43 +0000 (01:08 +0000)]
testsuite: extend timeout into all gm2 tests
Add timeout capability to gm2-torture.exp. Also add a simple
gm2_push_timeout/gm2_pop timeout facility and calibrate all
tests to use the default of 10 seconds. 15 seconds (for the
coroutine tests) and 60 seconds for whole program optimization.
gcc/testsuite/ChangeLog:
* gm2/coroutines/pim/run/pass/coroutines-pim-run-pass.exp
(timeout-dg.exp): Load.
Call gm2_push_timeout 15.
Call gm2_pop_timeout at the end.
* gm2/link/min/pass/link-min-pass.exp: Set path
argument to "".
* gm2/switches/whole-program/pass/run/switches-whole-program-pass-run.exp:
Call gm2_push_timeout 60.
Call gm2_pop_timeout at the end.
* lib/gm2-torture.exp (gm2_previous_timeout): Set to 10 or
individual_timeout.
Configure dejagnu to timeout for 10 seconds.
(gm2_push_timeout): New proc.
(gm2_pop_timeout): New proc.
* lib/gm2.exp (gm2_previous_timeout): Set to 10 or
individual_timeout.
Configure dejagnu to timeout for 10 seconds.
(gm2_push_timeout): New proc.
(gm2_pop_timeout): New proc.
Jonathan Wakely [Thu, 12 Jan 2023 22:35:30 +0000 (22:35 +0000)]
libstdc++: Fix exports for IEEE128 versions of __try_use_facet [PR108327]
The new symbols need to be exported, as well as some of the
std::locale::facet::id globals, which are not new but were presumably
not needed by any inline functions before now.
libstdc++-v3/ChangeLog:
PR libstdc++/108327
* config/os/gnu-linux/ldbl-extra.ver (GLIBCXX_LDBL_3.4.31):
Export __try_use_facet specializations for facets in namespace
__gnu_cxx_ldbl128.
* config/os/gnu-linux/ldbl-ieee128-extra.ver
(GLIBCXX_IEEE128_3.4.31): Likewise for facets in namespace
__gnu_cxx_ieee128.
* testsuite/util/testsuite_abi.cc: Add to lists of known and
latest versions.
Jonathan Wakely [Thu, 12 Jan 2023 13:03:01 +0000 (13:03 +0000)]
libstdc++: Do not include <system_error> in concurrency headers
The <condition_variable>, <mutex>, and <shared_mutex> headers use
std::errc constants, but don't use std::system_error itself. They only
use the __throw_system_error(int) function, which is defined in
<bits/functexcept.h>.
By including the header for the errc constants instead of the whole of
<system_error> we avoid depending on the whole std::string definition.
Roger Sayle [Thu, 12 Jan 2023 21:47:40 +0000 (21:47 +0000)]
PR tree-optimization/92342: Optimize b & -(a==c) in match.pd
This patch is an update/tweak of Andrew Pinski's two patches for
PR tree-optimization/92342, that were originally posted by in November:
https://gcc.gnu.org/pipermail/gcc-patches/2021-November/585111.html
https://gcc.gnu.org/pipermail/gcc-patches/2021-November/585112.html
Technically, the first of those was approved by Richard Biener, though
never committed, and my first thought was to simply push it for Andrew,
but the review of the second piece expressed concerns over comparisons
in non-integral modes, where the result may not be zero-one valued.
Indeed both transformations misbehave in the presence of vector mode
comparisons (these transformations are already implemented for
vec_cond elsewhere in match.pd), so my minor contribution is to limit
these new transformations to scalars, by testing that both the operands
and results are INTEGRAL_TYPE_P.
2023-01-12 Andrew Pinski <apinski@marvell.com>
Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog:
PR tree-optimization/92342
* match.pd ((m1 CMP m2) * d -> (m1 CMP m2) ? d : 0):
Use tcc_comparison and :c for the multiply.
(b & -(a CMP c) -> (a CMP c)?b:0): New pattern.
gcc/testsuite/ChangeLog:
PR tree-optimization/92342
* gcc.dg/tree-ssa/andnegcmp-1.c: New test.
* gcc.dg/tree-ssa/andnegcmp-2.c: New test.
* gcc.dg/tree-ssa/multcmp-1.c: New test.
* gcc.dg/tree-ssa/multcmp-2.c: New test.
Christophe Lyon [Fri, 25 Nov 2022 13:35:11 +0000 (13:35 +0000)]
aarch64: Fix bit-field alignment in param passing [PR105549]
While working on enabling DFP for AArch64, I noticed new failures in
gcc.dg/compat/struct-layout-1.exp (t028) which were not actually
caused by DFP types handling. These tests are generated during 'make
check' and enabling DFP made generation different (not sure if new
non-DFP tests are generated, or if existing ones are generated
differently, the tests in question are huge and difficult to compare).
Anyway, I reduced the problem to what I attach at the end of the new
gcc.target/aarch64/aapcs64/va_arg-17.c test and rewrote it in the same
scheme as other va_arg* AArch64 tests. Richard Sandiford further
reduced this to a non-vararg function, added as a second testcase.
This is a tough case mixing bit-fields and alignment, where
aarch64_function_arg_alignment did not follow what its descriptive
comment says: we want to use the natural alignment of the bit-field
type only if the user didn't reduce the alignment for the bit-field
itself.
The patch also adds a comment and assert that would help someone who
has to look at this area again.
The fix would be very small, except that this introduces a new ABI
break, and we have to warn about that. Since this actually fixes a
problem introduced in GCC 9.1, we keep the old computation to detect
when we now behave differently.
This patch adds two new tests (va_arg-17.c and
pr105549.c). va_arg-17.c contains the reduced offending testcase from
struct-layout-1.exp for reference. We update some tests introduced by
the previous patch, where parameters with bit-fields and packed
attribute now emit a different warning.
2022-11-28 Christophe Lyon <christophe.lyon@arm.com>
Richard Sandiford <richard.sandiford@arm.com>
gcc/
PR target/105549
* config/aarch64/aarch64.cc (aarch64_function_arg_alignment):
Check DECL_PACKED for bitfield.
(aarch64_layout_arg): Warn when parameter passing ABI changes.
(aarch64_function_arg_boundary): Do not warn here.
(aarch64_gimplify_va_arg_expr): Warn when parameter passing ABI
changes.
Christophe Lyon [Tue, 14 Jun 2022 21:08:33 +0000 (21:08 +0000)]
aarch64: fix warning emission for ABI break since GCC 9.1
While looking at PR 105549, which is about fixing the ABI break
introduced in GCC 9.1 in parameter alignment with bit-fields, we
noticed that the GCC 9.1 warning is not emitted in all the cases where
it should be. This patch fixes that and the next patch in the series
fixes the GCC 9.1 break.
We split this into two patches since patch #2 introduces a new ABI
break starting with GCC 13.1. This way, patch #1 can be back-ported
to release branches if needed to fix the GCC 9.1 warning issue.
The main idea is to add a new global boolean that indicates whether
we're expanding the start of a function, so that aarch64_layout_arg
can emit warnings for callees as well as callers. This removes the
need for aarch64_function_arg_boundary to warn (with its incomplete
information). However, in the first patch there are still cases where
we emit warnings were we should not; this is fixed in patch #2 where
we can distinguish between GCC 9.1 and GCC.13.1 ABI breaks properly.
The fix in aarch64_function_arg_boundary (replacing & with &&) looks
like an oversight of a previous commit in this area which changed
'abi_break' from a boolean to an integer.
We also take the opportunity to fix the comment above
aarch64_function_arg_alignment since the value of the abi_break
parameter was changed in a previous commit, no longer matching the
description.
2022-11-28 Christophe Lyon <christophe.lyon@arm.com>
Richard Sandiford <richard.sandiford@arm.com>
* gcc.target/aarch64/bitfield-abi-warning-align16-O2.c: New test.
* gcc.target/aarch64/bitfield-abi-warning-align16-O2-extra.c: New
test.
* gcc.target/aarch64/bitfield-abi-warning-align32-O2.c: New test.
* gcc.target/aarch64/bitfield-abi-warning-align32-O2-extra.c: New
test.
* gcc.target/aarch64/bitfield-abi-warning-align8-O2.c: New test.
* gcc.target/aarch64/bitfield-abi-warning.h: New test.
* g++.target/aarch64/bitfield-abi-warning-align16-O2.C: New test.
* g++.target/aarch64/bitfield-abi-warning-align16-O2-extra.C: New
test.
* g++.target/aarch64/bitfield-abi-warning-align32-O2.C: New test.
* g++.target/aarch64/bitfield-abi-warning-align32-O2-extra.C: New
test.
* g++.target/aarch64/bitfield-abi-warning-align8-O2.C: New test.
* g++.target/aarch64/bitfield-abi-warning.h: New test.
Richard Biener [Thu, 12 Jan 2023 10:18:22 +0000 (11:18 +0100)]
tree-optimization/99412 - reassoc and reduction chains
With -ffast-math we end up associating reduction chains and break
them - this is because of old code that tries to rectify reductions
into a shape likened by the vectorizer. Nowadays the rank compute
produces correct association for reduction chains and the vectorizer
has robust support to fall back to a regular reductions (via
reduction path) when it turns out to be not a proper reduction chain.
So this patch removes the special code in reassoc which makes
the TSVC s352 vectorized with -Ofast (it is already without
-ffast-math).
This branch instruction has short encoding if EQ/NE comparison against
immediate zero when the Code Density Option is enabled, but its "length"
attribute was only for normal encoding. This patch fixes it.
This patch also prevents undesireable replacement the comparison immediate
zero of the instruction (short encoding, as mentioned above) with a
register that has value of zero (normal encoding) by the postreload pass.
gcc/ChangeLog:
* config/xtensa/xtensa.md (*btrue):
Correct value of the attribute "length" that depends on
TARGET_DENSITY and operands, and add '?' character to the register
constraint of the compared operand.
Jonathan Wakely [Thu, 5 Jan 2023 20:23:26 +0000 (20:23 +0000)]
libstdc++: Use lock-free type for __platform_wait_t
For non-futex targets the __platform_wait_t type is currently uint64_t,
but that requires a lock in libatomic for some 32-bit targets. We don't
really need a 64-bit type, so use unsigned long if that is lock-free,
and int otherwise. This should mean it's lock-free on a wider set of
targets.
libstdc++-v3/ChangeLog:
* include/bits/atomic_wait.h (__detail::__platform_wait_t):
Define as unsigned long if always lock-free, and unsigned int
otherwise.
Lewis Hyatt [Thu, 29 Dec 2022 21:55:21 +0000 (16:55 -0500)]
preprocessor: Don't register pragmas in directives-only mode [PR108244]
libcpp's directives-only mode does not expect deferred pragmas to be
registered, but to date the c-family registration process has not checked for
this case. That issue became more visible since r13-1544, which added the
commonly used GCC diagnostic pragmas to the set of those registered in
preprocessing modes. Fix it by checking for directives-only mode in
c-family/c-pragma.cc.
gcc/c-family/ChangeLog:
PR preprocessor/108244
* c-pragma.cc (c_register_pragma_1): Don't attempt to register any
deferred pragmas if -fdirectives-only.
(init_pragma): Likewise.
gcc/testsuite/ChangeLog:
* c-c++-common/cpp/pr108244-1.c: New test.
* c-c++-common/cpp/pr108244-2.c: New test.
* c-c++-common/gomp/pr108244-3.c: New test.
David Malcolm [Wed, 11 Jan 2023 21:27:06 +0000 (16:27 -0500)]
analyzer: fix leak false positives on "*UNKNOWN = PTR;" [PR108252]
PR analyzer/108252 reports a false positive from -Wanalyzer-malloc-leak on
code like this:
*ptr_ptr = strdup(EXPR);
where ptr_ptr is an UNKNOWN_VALUE.
When we handle:
*UNKNOWN = PTR;
store::set_value normally marks *PTR as having escaped, and this means
we don't report PTR as leaking when the last usage of PTR is lost.
However this only works for cases where PTR is a region_svalue.
In the example in the bug, it's a conjured_svalue, rather than a
region_svalue. A similar problem can arise for FDs, which aren't
pointers.
This patch fixes the bug by updating store::set_value to mark any
values stored via *UNKNOWN = VAL as not leaking.
Additionally, sm-malloc.cc's known_allocator_p hardcodes strdup and
strndup as allocators (and thus transitioning their result to
"unchecked"), but we don't implement known_functions for these, leading
to the LHS being a CONJURED_SVALUE, rather than a region_svalue to a
heap-allocated region. A similar issue happens with functions marked
with __attribute__((malloc)). As part of a "belt and braces" fix, the
patch also updates the handling of these functions, so that they use
heap-allocated regions.
gcc/analyzer/ChangeLog:
PR analyzer/108252
* kf.cc (class kf_strdup): New.
(class kf_strndup): New.
(register_known_functions): Register them.
* region-model.cc (region_model::on_call_pre): Use
&HEAP_ALLOCATED_REGION for the default result of an external
function with the "malloc" attribute, rather than CONJURED_SVALUE.
(region_model::get_or_create_region_for_heap_alloc): Allow
"size_in_bytes" to be NULL.
* store.cc (store::set_value): When handling *UNKNOWN = VAL,
mark VAL as "maybe bound".
gcc/testsuite/ChangeLog:
PR analyzer/108252
* gcc.dg/analyzer/attr-malloc-pr108252.c: New test.
* gcc.dg/analyzer/fd-leak-pr108252.c: New test.
* gcc.dg/analyzer/flex-with-call-summaries.c: Remove xfail from
warning false +ve directives.
* gcc.dg/analyzer/pr103217-2.c: Add -Wno-analyzer-too-complex.
* gcc.dg/analyzer/pr103217-3.c: Likewise.
* gcc.dg/analyzer/strdup-pr108252.c: New test.
* gcc.dg/analyzer/strndup-pr108252.c: New test.
Signed-off-by: David Malcolm <dmalcolm@redhat.com>
Jakub Jelinek [Wed, 11 Jan 2023 21:18:42 +0000 (22:18 +0100)]
c: Don't emit DEBUG_BEGIN_STMTs for K&R function argument declarations [PR105972]
K&R function parameter declarations are handled by calling
recursively c_parser_declaration_or_fndef in a loop, where each such
call will add_debug_begin_stmt at the start.
Now, if the K&R function definition is not a nested function,
building_stmt_list_p () is false and so we don't emit the DEBUG_BEGIN_STMTs
anywhere, but if it is a nested function, we emit it in the containing
function at the point of the nested function definition.
As the following testcase shows, it can cause ICEs if the containing
function has var-tracking disabled but nested function has them enabled,
as the DEBUG_BEGIN_STMTs are added to the containing function which
shouldn't have them but MAY_HAVE_DEBUG_MARKER_STMTS is checked already
for the nested function, or just wrong experience in the debugger.
The following patch ensures we don't emit any such DEBUG_BEGIN_STMTs for the
K&R function parameter declarations even in nested functions.
2023-01-11 Jakub Jelinek <jakub@redhat.com>
PR c/105972
* c-parser.cc (c_parser_declaration_or_fndef): Disable debug non-bind
markers for K&R function parameter declarations of nested functions.
Roger Sayle [Wed, 11 Jan 2023 16:54:58 +0000 (16:54 +0000)]
PR tree-optimization/71343: Value number X<<2 as X*4.
This patch is the second part of a fix for PR tree-optimization/71343,
that implements Richard Biener's suggestion of using tree-ssa's value
numbering instead of match.pd. The change is that when assigning a
value number for the expression X<<C, we actually look-up or insert
the value number for the multiplication X*(1<<C). This elegantly
handles the fact that we (intentionally) don't canonicalize these as
equivalent in GIMPLE, and the optimization/equivalence in PR 71343 now
happens by (tree-ssa SCCVN) magic.
2023-01-11 Roger Sayle <roger@nextmovesoftware.com>
gcc/ChangeLog
PR tree-optimization/71343
* tree-ssa-sccvn.cc (visit_nary_op) <case LSHIFT_EXPR>: Make
the value number of the expression X << C the same as the value
number for the multiplication X * (1<<C).
gcc/testsuite/ChangeLog
PR tree-optimization/71343
* gcc.dg/pr71343-2.c: New test case.
David Faust [Tue, 10 Jan 2023 18:53:12 +0000 (10:53 -0800)]
bpf: correct bpf_print_operand for floats [PR108293]
The existing logic in bpf_print_operand was only correct for integral
CONST_DOUBLEs, and emitted garbage for floating point modes. Fix it so
floating point mode operands are correctly handled.
PR target/108293
gcc/
* config/bpf/bpf.cc (bpf_print_operand): Correct handling for
floating point modes.
gcc/testsuite/
* gcc.target/bpf/double-1.c: New test.
* gcc.target/bpf/double-2.c: New test.
* gcc.target/bpf/float-1.c: New test.
Kewen Lin [Wed, 11 Jan 2023 12:59:24 +0000 (06:59 -0600)]
rs6000/test: Make ppc-fortran.exp only available for PowerPC target
When testing one patch which adds a fortran test case into
test bucket powerpc/ppc-fortran/, I found one unexpected
failure on a non-PowerPC target. It's due to that
ppc-fortran.exp does not exit early if the testing target
isn't a PowerPC target. This patch is to make it exit
immediately if the testing target isn't a PowerPC target.
gcc/testsuite/ChangeLog:
* gcc.target/powerpc/ppc-fortran/ppc-fortran.exp: Exit immediately if
the testing target isn't a PowerPC target.
Kewen Lin [Wed, 11 Jan 2023 12:59:08 +0000 (06:59 -0600)]
rs6000: Make P10_FUSION honour tuning setting
We noticed this issue when Segher reviewed the patch for
PR104024. When there is no explicit setting for option
-mpower10-fusion, we enable OPTION_MASK_P10_FUSION for
TARGET_POWER10. But it's not right, it should honour
tuning setting instead.
This patch is to fix it accordingly, it's bootstrapped
, and regtested on powerpc64-linux-gnu P8 and
powerpc64le-linux-gnu P9.
But on powerpc64le-linux-gnu P10 it had one regression
failure against the test case gcc.target/powerpc/pr105586.c.
I looked into it and confirmed that a latent bug was
exposed and filed one separated bug PR108273 instead.
Richard Biener [Mon, 9 Jan 2023 08:42:22 +0000 (09:42 +0100)]
tree-optimization/107767 - not profitable switch conversion
When the CFG has not merged equal PHI defs in a switch stmt the
cost model from switch conversion gets off and we prefer a
jump table over branches. The following fixes that by recording
cases that will be merged later and more appropriately counting
unique values.
Jakub Jelinek [Wed, 11 Jan 2023 12:06:14 +0000 (13:06 +0100)]
testsuite: Enable pr108308.c test on all int32 targets [PR108308]
This test seems to rely on 32-bit int (and uses a wider constant
which shouldn't fit into int), I've initially enabled it on ilp32+lp64
target, but apparently it works on llp64 too, so I've changed it to
int32.
2023-01-11 Jakub Jelinek <jakub@redhat.com>
PR target/108308
* gcc.dg/pr108308.c: Use int32 target rather than { ilp32 || lp64 }.
Martin Liska [Wed, 28 Dec 2022 08:11:40 +0000 (09:11 +0100)]
switch expansion: limit JT growth param values
Currently, one can request a huge jump table creation which
leads to a non-sensual huge output. Moreover, use auto_vec rather
than a stack-allocated array.
PR middle-end/107976
gcc/ChangeLog:
* params.opt: Limit JT params.
* stmt.cc (emit_case_dispatch_table): Use auto_vec.
The following relaxes a heuristic that prevents creating irreducible
loops from FSM threads not covering multi-way branches. Instead of
allowing threads that adhere to
with reasoning "We also consider it worth creating an irreducible inner loop if
the number of copied statement is low relative to the length of the path --
in that case there's little the traditional loop optimizer would have done
anyway, so an irreducible loop is not so bad." that I cannot make much
sense of the following patch changes that to only allow those after
loop optimization and when they are (scaled) short:
This allows us to get rid of --param fsm-scale-path-blocks which
previous to the bisected revision allowed an enlarged path covering
the original allowance (but we do not consider that enlarged path
now because enlarging it doesn't add any information).